Most AI-related threats can be prevented without altering the core network protocol, the developer believes. To avoid hacker attacks on the blockchain, client software, or mining infrastructure, it is far simpler to deploy updates than to urgently change Bitcoin’s network rules or transaction history, Buterin stated. According to him, hackers can use AI for phishing, stealing credentials, developing malware, and finding vulnerabilities in code; however, this technology can also be used for beneficial purposes, such as analyzing code, testing updates, and detecting suspicious activity.
The probability of a real hack of the hashes or the Proof-of-Work (PoW) consensus algorithm upon which Bitcoin operates is negligible, says the programmer. Bitcoin mining uses SHA-256—a hash function that proves miners have performed computational work. A conventional hack would require attackers to find a vulnerability that drastically reduces the effort needed to generate valid hashes. However, current generative AI systems do not yet possess such capabilities—they can improve software analysis and vulnerability detection, but cannot automatically break existing cryptographic functions, the crypto entrepreneur notes.
I take the opposite side of that.
My basic reasons are that I am quite optimistic about cybersecurity in the long term and I see the primary problem as being the transition, and I expect BTC to handle at least any issues that do not require social consensus well…
— vitalik.eth (@VitalikButerin) September 7, 2026Vitalik Buterin made his comment during a discussion with Liron Shapira, host of the Doom Debates podcast. The blogger, with approximately 40,600 followers on social media X, is known for publicly predicting that artificial intelligence poses a threat to human existence. Shapira asserts that AI could be used to hack Bitcoin. He fears that due to this threat, Bitcoin could crash by 50% within the next two years.
Previously, Buterin stated that the main threat of AI is not in the evolving intelligence of machines, but in the desire of corporations and governments to gain their own unilateral control over neural networks.